computational tools for univariate and multivariate GWAS using quantitative traits

Question

[fastDio]

Hi all,

I have just started learning GWAS as I will have to run some of them in the near future and I am pretty confused about which are the best computational tools available for my dataset. In particular, I have almost 10 millions SNPs that I would like to test against clinical numerical covariates such as blood pressure. Moreover, I have some high-dimensional phenotypes that I would like to reduce to 40/50 variables using PCA or ICA and test them in the same multivariate model.

I have been advised to use the R package matrix EQTL for univariate (linear regression) models and from their paper the GEMMA software (http://www.xzlab.org/software.html) looks promising for multivariate ones. However, I would like to year your thoughts about this before starting, please.

Many thanks!

Answer 1

[ddiez]

I am not an expert on GWAS but if you want to do this in R I would take a look at Bioconductor. In particular the packages related to GWAS. Also search for QTL or other terms of interest in the view of available packages.

Answer 2

[anp375]

What about PLINK? Is the number of variables too much for PLINK?

[chrchang523]

PLINK currently only supports univariate models, and it is not optimized for a huge number of phenotypes in the way Matrix eQTL is.

[anp375]

Thank you for the explanation. This will be helpful.